Berlin 2024 – wissenschaftliches Programm

Bereiche | Tage | Auswahl | Suche | Aktualisierungen | Downloads | Hilfe

SOE: Fachverband Physik sozio-ökonomischer Systeme

SOE 2: Focus Session: Machine Learning for Complex Socio-economic Systems

SOE 2.2: Vortrag

Montag, 18. März 2024, 10:00–10:15, MA 001

Collaboration, not polarization: A Relational Graph Convolutional Network (RGCN) model to disentangle active and passive cosponsorship in the U.S. Congress — •Frank Schweitzer — Chair of Systems Design, ETH Zurich, Switzerland

Public coverage fuels the impression of increasing elite polarization and paralysis in the U.S. Congress. The other half of the truth is the fact that, e.g., more than 15.000 bills were introduced to the 115th U.S. Congress (2017-2019). Legislators from both parties cosponsor these bills actively, e.g. by drafting the bill, or passively, by their signature. To identify their motivation, we have curated and analyzed a large data set containing bill texts, legislator speeches, and cosponsorship data for all bills from the 112th to 115th U.S. Congress. We use Natural Language Processing to obtain contextual embeddings of bills and speeches and to extract a citation network between legislators. We then develop and train a RGCN to predict active and passive cosponsorship relations. Our results demonstrate that the two types of cosponsorship are driven by two different motivations: the backing of political colleagues and the backing of the bill’s content.

Reference: G. Russo, C. Gote, L. Brandenberger, S. Schlosser, F. Schweitzer: Helping a Friend or Supporting a Cause? Disentangling Active and Passive Cosponsorship in the U.S. Congress, Proc. 61st Ann. Meeting Assoc. Comput. Linguistics, Vol. 1: Long Papers, pp. 2952-2969 (2023) doi:10.18653/v1/2023.acl-long.166